Tag: machine-learning

All the articles with the tag "machine-learning".

• Wrong on Purpose

  Published:Oct 24, 2026 |  at 04:35 PM

  Random strings find bugs no designer imagined. Why structureless input is the only honest question you can ask a system, four decades of evidence for it, and where noise stops working.

• The Idle Machine

  Published:Oct 24, 2026 |  at 02:22 AM

  Autoregressive decoding leaves almost all of an accelerator's arithmetic idle while it waits on memory bandwidth. Batching and speculative decoding fill that idle budget by the same mechanism, and the acceptance rate tells you what the mechanism is worth.

• Secondhand Scars

  Published:Sep 27, 2026 |  at 03:11 PM

  A corpus is a record of other people's scars, and that is why pretraining works so well and stalls at the frontier, where nobody has been hurt yet. A hypothesis about why the world only writes into a system through damage.